50 likes | 150 Vues
This draft proposes the RLCI approach for TCP to efficiently handle mobility and connectivity disruptions, signaling indications through a new TCP option. RLCI aims to enhance performance, adapt faster to new paths, and improve responsiveness in diverse networking scenarios. The document discusses the benefits, updates, next steps, and potential for experimental RFC consideration.
E N D
TCP Response to Lower-Layer Connectivity-Change Indications draft-schuetz-tcpm-tcp-rlci-01 Simon Schuetz NEC Lars Eggert Nokia Wesley Eddy Verizon Yogesh Swami, Khiem Le Nokia IETF-68, Prague, Czech Republic
Problem to Solve and Basic Idea • TCP is (mainly) unaware of events in lower layers can be inefficient in case of mobility and/or temporary connectivity disruptions • RLCI • uses generic (technology-independent) indications from lower-layers (“something happened”) • avoids long idle time due to (repetitive) RTOs • re-probes path state to adapt faster to changed conditions • signals local indications to peer hosts through a new TCP option IETF-68, Prague, Czech Republic
Why RLCI at TCPM? • RLCI provides generic approach to overcome problems in diverse networking scenarios • handovers • adapt faster to new paths • friendlier: avoid over-loading new path • performance: quickly use available capacity • connectivity disruptions • Increase performance and responsiveness IETF-68, Prague, Czech Republic
What’s new? • -00 version was presented at IETF-66 • cleaned up description of connectivity-change indication processing • specification more formal • developed preliminary NS-2 prototype • first simulation results look promising • repeated “long disruption” scenarios from Simon’s and Lars’ CCR paper on retransmit-now mechanism* • got similar (good) results on performance improvements • final simulation results will be made available * Protocol Enhancements for Intermittently Connected Hosts. Simon Schütz, Lars Eggert, Stefan Schmid and Marcus Brunner. ACM Computer Communication Review (CCR), Vol. 35, No. 3, July 2005, pp. 5-18. IETF-68, Prague, Czech Republic
Next Steps • Would like to start discussion on mailing list • Is the problem considered worth being followed up? • Is the RLCI approach the right way to go? • Are there technical issues with current draft? • Simulation results should be made available as soon as possible • Consider it as candidate for Experimental RFC IETF-68, Prague, Czech Republic